mcpField
D

DeFiMind

Remote
defimind.aiFinance0Updated 5d ago

About DeFiMind

DeFiMind (echoledger-mcp) is a decommissioned MCP server for analyzing live liquidity pools on Uniswap V2/V3, Balancer, and Curve. It offered pool and position analytics plus portable state snapshots for local counterfactual analysis; the hosted endpoint is offline and the project is not maintained.

Key features

  1. 01LP position PnL analysis
  2. 02Price-move and slippage simulation
  3. 03Pool health and rug-signal checks
  4. 04Curve depeg-risk analysis
  5. 05Portable State Twin generation

Use cases

  1. 01Review LP position performance
  2. 02Estimate trade slippage
  3. 03Check pool condition and rug signals
  4. 04Run local scenarios from a pool snapshot

MCP support

The server exposed 11 analytics tools for supported Uniswap V2/V3, Balancer, and Curve pools, using a caller-supplied RPC URL. Its public hosted endpoint has been taken offline.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://mcp.defimind.ai/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

Details

From the project README

⚠ Decommissioned (2026-10-23). This MCP server is no longer hosted. The public endpoint mcp.echoledger.ai has been taken offline and its listings removed. Kept for reference. Not maintained, not deployed.

echoledger-mcp

EchoLedger's MCP server. Analyze live Uniswap V2/V3, Balancer, and Curve stableswap pools — positions, price moves, pool health, rug signals, slippage, and depeg risk — or build a portable State Twin for off-MCP analysis, from any MCP-compatible AI client. Reads real chain state via a caller-supplied RPC; the endpoint itself is authless.

License: Apache-2.0 MCP Python arXiv smithery badge

Endpoint: https://mcp.echoledger.ai/mcp (streamable-HTTP, authless).

Install

Claude Desktop

Settings → Integrations → Add server. URL: https://mcp.echoledger.ai/mcp

Claude Code
claude mcp add --transport http echoledger https://mcp.echoledger.ai/mcp
Cursor

Settings → MCP → Add new MCP server. Name: echoledger · Type: http · URL: https://mcp.echoledger.ai/mcp

Smithery

Listed at smithery.ai/servers/ic3moore/echoledger — connect via the Smithery gateway from any supported client.

Tools

The endpoint ships 11 tools over live Uniswap V2/V3, Balancer weighted (2-asset), and Curve stableswap (2-asset) pools. Two surfaces:

  • 10 reactive primitives — one question, one answer, one chain read. Four of them also take a vector input (e.g. price_change_pcts[], amounts_in[]) to sweep a whole grid/curve in one call.
  • BuildStateTwin — returns a portable State Twin (the pool's state as JSON) that a client rehydrates locally to run unlimited counterfactuals off the MCP, zero further RPC.

Every call takes pool_address, rpc_url, and pool_type, plus an optional chain_id guard and block_number pin — nothing is stored. Each reactive tool is protocol-specific and advertises only the pool_type values it accepts; pointing one at an unsupported type returns a clean error before any chain read.

Full reference — params, returns, and a verified example per tool: docs/TOOLS.md.

Scope: the Balancer tools cover 2-asset weighted pools; the stableswap tools cover 2-asset plain Curve pools (rate-bearing pools — metapools, LSD — are a later release). A stableswap tool pointed at a 3-coin or rate-bearing pool fails cleanly.

Uniswap V2/V3
AnalyzePosition

PnL decomposition for an LP position — impermanent loss, fees, and net.

"Analyze my position in the V3 USDC/ETH pool."

SimulatePriceMove

"What if price moves X%?" — projected reserves, position value, and IL.

"If ETH drops 20%, what happens to my V3 ETH/DAI position?"

CheckPoolHealth

TVL, reserves, LP concentration, and recent activity for a pool.

"Is this V3 pool healthy?"

DetectRugSignals

Threshold-based rug flags on a pool's on-chain state.

"Any rug signals on this V2 pair?"

CalculateSlippage

Slippage, price impact, and max trade size for a given trade.

"How much slippage on a 50 ETH buy in this V3 pool?"

Balancer (2-asset weighted)
AnalyzeBalancerLP

PnL decomposition for a 2-asset Balancer weighted-pool position, using the weighted-pool IL formula where the token weight shapes IL magnitude.

"How is my 80/20 BAL/WETH Balancer position doing?"

SimulateBalancerMove

"What if the base token moves X%?" — projected value and IL on a 2-asset weighted pool, weight-aware.

"If BAL drops 30%, what happens to my BAL/WETH Balancer LP?"

Curve stableswap (2-asset plain)
AnalyzeStableswapLP

PnL decomposition for a 2-asset Curve stableswap position via the amplified-invariant IL formula — small depegs can produce outsized IL at high A.

"Analyze my position in the crvUSD/USDC pool."

SimulateStableswapMove

"What if the peg shifts X%?" — projected value and IL on a 2-asset stableswap pool. At high A, large shocks may be physically unreachable (returned as null).

"What happens to my USDC/DAI Curve LP if USDC depegs 2%?"

AssessDepegRisk

IL across a ladder of depeg levels (default 2%, 5%, 10%, 20%, 50%) for a 2-asset stableswap position, with an optional constant-product benchmark.

"How exposed is my crvUSD/USDC position to a depeg?"

State twin builder (all four pool types)
BuildStateTwin

Reads a pool once and returns a portable State Twin — the pool's state as JSON (__type__ + fields + a content_hash). A client rehydrates it locally and runs unlimited counterfactuals (price moves, IL, slippage) off the MCP, with zero further RPC — build once, run N. Spans all four pool types. See the twin round-trip in docs/TOOLS.md and the echoledger package.

"Build me a reusable twin of the USDC/ETH 0.05% pool."

(Honest gap: it's a single-block STATE twin — history-derived health metrics stay server-side reads inside CheckPoolHealth/DetectRugSignals.)

How it works

echoledger-mcp reads live Uniswap V2/V3, Balancer, and Curve stableswap pool state through a caller-supplied RPC endpoint (BYO-RPC, passed per tool call). The endpoint is authless — no API key, no account. Nothing is logged or stored; each call pulls state, runs the analysis, and returns a typed result.

The analytics are powered by open-source DeFiPy — the same primitives, twin, and tools that run against synthetic recipes, now pointed at real pools. The math is open; the reports are paid.

Roadmap

  • v0.2 — Balancer & Stableswap live reads ✓ shipped. The 5 Balancer/Stableswap tools went live on DeFiPy 2.2's LiveProviders (2-asset weighted + 2-asset plain Curve).
  • v0.2.2 — honest schemas, vectors & the twin builder ✓ shipped. Per-tool pool_type enums; vectorized scenario inputs on the four scenario tools; and BuildStateTwin — the 11th tool — for portable, off-MCP twin analysis.
  • N-asset & rate-bearing pools — 3-asset Balancer, N-coin and rate-bearing (metapool/LSD) Curve, once the upstream DeFiPy primitives extend past 2 assets.
  • Distribution polish — OCI-package listing, DNS-verified namespace, and OIDC auto-publish.

Develop

python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"

Run the server over streamable HTTP (binds to $PORT, default 8080):

PORT=8080 python -m echoledger_mcp.server      # endpoint at http://localhost:8080/mcp

Or over stdio for a local MCP Inspector smoke-test:

python -m echoledger_mcp.server --stdio
npx @modelcontextprotocol/inspector python -m echoledger_mcp.server --stdio

Tests (offline — fake provider + real twin/primitives):

pytest tests/

Live-RPC gate (real mainnet Uniswap V2/V3, Balancer, and Curve pools; needs your own RPC):

ECHOLEDGER_TEST_RPC_URL="https://eth-mainnet.example/v2/<key>" pytest tests/test_live.py -v

License

Apache-2.0. See LICENSE and NOTICE.

This project adapts code from DeFiPy (Apache-2.0).

See also

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.